AI boom vs bubble? $37-billion enterprise surge shows decisive tilt, says Menlo Ventures

Representative image.

For all the speculation around over-investment and bubbly valuations, new data from Menlo Ventures suggests the enterprise artificial intelligence (AI) market is expanding with real demand behind it.

Enterprise AI spending has surged to $37 billion in 2025, making it the fastest-scaling software category in history, the firm has said. Anthropic, which is backed by Menlo, has overtaken OpenAI to command an estimated 40 percent of enterprise LLM spend, helped by its lead in coding use cases.

The findings are based on a survey of 495 US enterprise AI decision-makers combined with a bottom-up model of spending across applications, infrastructure and model APIs.

The result is a clear signal that the market is in a boom rather than drifting toward a bubble, Menlo has claimed.

“An MIT study claiming that 95 percent of Gen AI initiatives fail rattled markets over the summer, exposing how quickly sentiment could shift beneath the weight of AI’s massive capex spend. The concerns aren’t unfounded, given the magnitude of the numbers. But the demand side tells a different story: Our latest market data shows broad adoption, real revenue, and productivity gains at scale, signaling a boom versus a bubble,” a blogpost by Menlo Ventures’ Tim Tully, Joff Redfern, Deedy Das and Derek Xiao said.

Demand outpaces the hype cycle

Menlo’s report said enterprise AI spend tripled from $11.5 billion in 2024, defying concerns around studies that claimed high failure rates for Generative AI (Gen AI) projects.

Instead of cooling, adoption is widening across functions such as coding, sales, customer service, healthcare, and horizontal copilots.

Application-layer tools captured $19 billion, or more than half of all Gen AI spend, and now represent over six percent of the global software market.

Source: Menlo Ventures, Gartner

Buying, not building, is fuelling rapid scale

A year ago, enterprises were evenly split between building and buying AI solutions.

Today, 76 percent of AI use cases are purchased, not built, as companies gravitate toward ready-made tools to deliver immediate productivity gains.

It said the shift runs against the bubble narratives because high-intent buyers are moving solutions into production faster than traditional software: AI deals convert at 47 percent, nearly double of SaaS norms.

Source: Menlo Ventures

Bottom-up adoption accelerates the boom

The strongest evidence against a bubble is the speed at which AI spreads inside enterprises without top-down push, Menlo said.

Product-led growth (PLG) now accounts for 27 percent of AI application spend, and shadow adoption could push that figure near 40 percent.

For example, tools like Cursor, n8n, ElevenLabs, and Wispr Flow are being used by individual employees long before procurement teams get involved.

Startups for speed; incumbents win deeper infrastructure

Startups now capture nearly two dollars for every one dollar earned by incumbents across AI applications, on the back of rapid iteration and feature rollout.

Coding tools, in particular, have become AI’s first true breakout category.

At the same time, the infrastructure layer remains more balanced: incumbents such as Databricks, Snowflake, MongoDB, and Datadog retain a 56 percent share as enterprises continue to rely on established data platforms.

Anthropic extends its enterprise lead

The model landscape has shifted sharply as well.

Anthropic has overtaken OpenAI to command an estimated 40 percent of enterprise LLM spend, helped by its lead in coding use cases.

Google’s Gemini models have also gained share, while enterprise adoption of open-source LLMs has fallen to 11 percent despite rising popularity among developers outside the enterprise.

Menlo cites this concentration of spend around proven models as another marker of a maturing, not speculative, market.

Infrastructure spend doubles as deployments deepen

Enterprise spending on AI infrastructure doubled to $18 billion in 2025 across foundation model APIs, training systems and orchestration tools.

Despite the hype around agentic AI, Menlo finds that only a small fraction of deployments use true agent architectures.

This shows that enterprises are scaling AI pragmatically rather than chasing untested abstractions.

A boom supported by revenue, adoption, and intent

Menlo Ventures believes that the numbers reflect a decisive tilt toward a boom: enterprises are expanding budgets, converting high-intent deals at pace, and generating meaningful productivity improvements.

With Anthropic’s model gains, PLG-driven scale, and accelerating use cases across sectors like healthcare and finance, Menlo expects 2026 to deepen AI’s real-world impact rather than trigger a correction.